DenseBox: Unifying Landmark Localization with End to End Object Detection

By Lichao Huang, Yi Yang, Yafeng Deng, Yinan Yu. 2015

484 citations recorded by the source corpus.

How can a single fully convolutional neural network (FCN) perform on object detection? We introduce DenseBox, a unified end-to-end FCN framework that directly predicts bounding boxes and object class confidences through all locations and scales of an image. Our contribution is two-fold. First, we show that a single FCN, if designed and optimized carefully, can detect multiple different objects extremely accurately and efficiently. Second, we show that when incorporating with landmark localization during multi-task learning, DenseBox further improves object detection accuray. We present experimental results on public benchmark datasets including MALF face detection and KITTI car detection, that indicate our DenseBox is the state-of-the-art system for detecting challenging objects such as faces and cars.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, methodology): “DenseBox [20] uses a filled circle located in the center of the object to define positives and then predicts the four distances from positives to the bound of the object bounding box for location.”
FCOS: Fully Convolutional One-Stage Object Detection (2019, methodology): “However, to handle the bounding boxes with different sizes, DenseBox [12] crops and resizes training images to a fixed scale.”
CoupleNet: Coupling Global Structure with Local Parts for Object Detection (2017, methodology): “Although there are end-to-end detection frameworks, like SSD [18], YOLO [20] and DenseBox [13], region-based systems ( i.e .”
3D fully convolutional network for vehicle detection in point cloud (2016, methodology): “Similar to DenseBox, the objectness region V is denoted as the center region of the object.”
Vehicle Detection from 3D Lidar Using Fully Convolutional Network (2016, methodology): “The trunk part of the proposed CNN architecture is similar to Huang et al. [11], Long et al. [18].”
Acquisition of Localization Confidence for Accurate Object Detection (2018, methodology): “[14, 32] utilize anchor-free, fully-convolutional networks to handle object scale variation.”
Orientation-boosted Voxel Nets for 3D Object Recognition (2016, methodology): “[15] approach the task with a multi-tasking method.”
Mask Scoring R-CNN (2019, background): “Deep networks are dramatically driving the development of computer vision, leading to a series of state-of-the-art in tasks including classiﬁcation [22, 16, 35], object detection [12, 17, 32, 27, 33, 34], semantic segmentation [28, 4, 37, 18] etc .”
Region Proposal by Guided Anchoring (2019, methodology): “(iii) Anchor-free methods [14, 15, 25] usually have simple pipelines and produce final detection results within a single stage.”
Feature Selective Anchor-Free Module for Single-Shot Object Detection (2019, background): “DenseBox [15] first proposed a unified end-to-end fully convolutional framework that directly predicted bounding boxes.”
A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection (2017, methodology): “Building on the sliding-window paradigm of the Overfeat detector, other computationally-efficient approaches have emerged such as YOLO [27], SSD [19] and DenseBox [12].”
PIXOR: Real-time 3D Object Detection from Point Clouds (2018, methodology): “For single-class object detection, DenseBox [17] and EAST [38] show that single-stage detector also works well without using manually designed anchors.”
OTA: Optimal Transport Assignment for Object Detection (2021, background): “Without anchor box, FCOS [38], Foveabox [17] and their precursors [30, 14, 46] directly assign anchor points around the center of objects as positive samples, showing promising detection performance.”
SiamFC++: Towards Robust and Accurate Visual Tracking with Target Estimation Guidelines (2019, methodology): “Researchers have tried various ways to design anchor-free detectors, like predicting bounding boxes at points near the center of objects (Redmon et al. 2016; Huang et al. 2015), or detecting and grouping a pair of corners of a bounding box (Law and Deng 2018).”
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection (2021, background): “In contrast, anchor-free methods [12, 25, 31, 15, 37] do not need these prior settings and are thus neater with better universality.”
VarifocalNet: An IoU-aware Dense Object Detector (2020, methodology): “The popular methods include DenseBox [24], FASF [25], Fove-aBox [13], FCOS [9], ATSS [26] and SPAD [27].”
Multi-oriented Scene Text Detection via Corner Localization and Region Segmentation (2018, methodology): “The ﬁrst branch is based on general object detectors (SSD [30], YOLO [37] and DenseBox [18]), such as TextBoxes [27], FCRN [14] and EAST [53] etc. , which predict candidate bounding boxes directly.”
The YOLO Framework: A Comprehensive Review of Evolution, Applications, and Benchmarks in Object Detection (2024, unknown): “probabilities simultaneously [36]. detection by introducing a grid-based This innovative design facilitates a highly efficient detection process, making YOLO particularly suitable for applications needing real-time performance. approach to predict bounding boxes and class probabilities…”
TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes (2018, background): “Numerous innovative works [39, 29, 47, 17, 10, 46, 30, 36, 6, 9, 34, 22, 28, 40, 31] are proposed, achieving excellent performances on standard benchmarks.”